For medical content, Google rewards consensus — even when consensus is wrong

For medical content, Google rewards consensus — even when consensus is wrong

The question: does the Quality Rater Guidelines reward content that is correct, or content that aligns with established expert consensus?

For high-harm YMYL, consensus. The QRG instructs raters that medical content should align with well-established expert and scientific consensus, and that content contradicting consensus on YMYL health topics should generally be treated as low quality. This is a deliberate epistemic choice: at the scale of web search, raters and systems cannot adjudicate frontier scientific disputes, so they anchor to the current consensus of recognized bodies.

The uncomfortable implication is honest to state: consensus is sometimes wrong, and it always lags genuine discovery. A correct finding that has not yet become consensus will be rated lower than the prevailing view it overturns. The framework knowingly trades some accuracy at the frontier for safety at scale — because the cost of amplifying confident wrong medical claims to millions is judged higher than the cost of lagging the occasional correct heterodoxy.

The evidence: the guidelines repeatedly defer to authoritative medical and scientific institutions as the reference for YMYL health, rather than asking raters to evaluate primary evidence themselves. Raters are not equipped to be peer reviewers, and the guidelines don't pretend they are.

Counter-evidence and caveat: this is rater instruction; how faithfully systems implement consensus-alignment at scale is unproven, and 'consensus' is itself fuzzy for evolving topics. The mechanism by which a system identifies consensus is not public.

What we still don't know: how the systems handle topics where consensus is genuinely contested or shifting — the messy middle where institutions disagree. That is precisely where the consensus-anchor strategy is weakest, and where its failures, when they come, will be most visible.